Implementing a filesystem

Based on inputs from #143, I was curious about what could be the best way to implement a hierarchical filesystem with folders and documents. I started with a "flat" implementation where each file/folder is a document with a required field "path" which is its absolute path. The client would then get all the files and build a file tree.
The problem with this approach is when it comes to renaming: renaming a folder means renaming all the contained files/subfolders. One way to overcome this problem would be to have just one root document representing the filesystem and each file/folder is stored in a json array inside this document. Each folder would then have a "contents" array where all the contents are stored, recursively. The question here is: could this be a problem? In the case of a big number of files, the document would end up being huge.

Start by reading issue #143 and comparing the proposed flat path-based documents with the single recursive JSON document described here. No implementation files or tests are named; this issue is complete only after the filesystem representation and its behavior for large numbers of files are defined.